harlowja | looks like python2.6 is busted again | 03:34 |
---|---|---|
harlowja | lol | 03:34 |
harlowja | https://gist.github.com/harlowja/7e1e7106e81f5aa809645d76dd74bddd | 03:37 |
prometheanfire | meh, no one really uses that :P | 03:41 |
prometheanfire | is that still maintained even for sec bugs? | 03:41 |
harlowja | python 2.6, its prob not | 03:41 |
prometheanfire | :P | 03:41 |
harlowja | sadly cent6 is still maintained | 03:41 |
harlowja | and cloud-init 0.7.x still says 2.6 supported (i'll have to support it no matter, ha) | 03:42 |
harlowja | simple stuff like data = {v: "UNAVAILABLE" for v in fmap.values()} though isn't hard to fix | 03:43 |
harlowja | just a PITA that its not getting detected earlier | 03:43 |
harlowja | oddly though even 2.7 seems currently busted @ https://gist.github.com/harlowja/3ea5fd7ad3739704e3298ab35ceefd1c | 03:44 |
harlowja | thats on rhel7 | 03:44 |
ffledgling | smoser: I figured out the problem, it was in my yaml :| | 08:06 |
ffledgling | Sorry for all the trouble | 08:06 |
ffledgling | I also did change the qemu-kvm boot options and added a `-boot dc` to elimnate the race condition | 08:06 |
smoser | ffledgling, well, fwiw, -boot dc wont really fix any race. | 11:42 |
smoser | the race happens because cloud-init doesn't attempt to wait for disks that aren't there to become attached (as they may never become attached) | 11:43 |
smoser | and by not waiting, linux might not have enumerated them yet. | 11:43 |
ffledgling | smoser: I see | 11:54 |
smoser | as i said though, i've never seen it in practice. | 11:54 |
ffledgling | smoser: got it | 11:57 |
ffledgling | I really ended up linking cloud init once I got it working correctly. It was just strange when I started with it | 11:58 |
smoser | ffledgling, thanks. | 12:03 |
smoser | yaml is indeed easy to mis-type | 12:03 |
smoser | one very simple thing you can do is just: | 12:04 |
smoser | python -c 'import yaml, sys, pprint; pprint.pprint(yaml.load(open(sys.argv[1])))' | 12:04 |
smoser | and if it stack traces, thats bad yaml | 12:04 |
smoser | it doesn't do any further checking obviously, and we want to better lint or schema data, but do not at the moment | 12:05 |
ffledgling | smoser: Yeah, I think my problem was I was using a makefile to generate the yaml on the fly and I accidently did the thing where I converted tabs to spaces, which resulted in incorrect indentation. | 12:11 |
ffledgling | I think python on the whole doesn't have great support for YAML | 12:11 |
smoser | ? | 12:14 |
smoser | it has really good support for yaml | 12:14 |
ffledgling | in terms of identifying and point out syntax errors I mean | 12:14 |
smoser | tabs and spaces are by yaml specification not intermixable | 12:14 |
smoser | yeah, the loader doesn't give you a ton of good info. | 12:15 |
smoser | one thing you *can* do is just use json | 12:15 |
smoser | yaml is a strict superset of json | 12:15 |
ffledgling | the builtin json module is pretty bad as well, and very unforgiving iirc (simplejson is better I think?) | 12:15 |
ffledgling | Does cloud init support json? | 12:15 |
=== mgagne_ is now known as mgagne | ||
=== rtheis_ is now known as rtheis | ||
ffledgling | Is automatic package installation a new feature (post 0.7.6) ? For whatever reason the packages directive doesn't work for me | 13:56 |
Odd_Bloke | ffledgling: JSON is valid YAML, I believe, so yes. :) | 13:59 |
ffledgling | Ignore my old complaint about installation not working, it was just taking time and running in the background | 14:32 |
smoser | Odd_Bloke, yes. YAML is 100% strict superset of JSON | 14:45 |
smoser | its basically json with some nice human features like: | 14:45 |
smoser | a.) "anchors" which are like references and really nice | 14:46 |
smoser | b.) easily mis-typable in non-obvious ways | 14:46 |
smoser | magicalChicken, if you're lookin for things... on cloud-init cleanup | 16:31 |
smoser | one thing that would be nice.. in nplaces we stack trace and put a failure in the log (at warn level) | 16:31 |
smoser | but the strings are all one line and very hard to read | 16:32 |
smoser | it'd be nice if put well formatted human readable stack traces somewhere on failure. | 16:32 |
magicalChicken | smoser: right yeah that makes sense | 16:37 |
magicalChicken | similar to the issue with curtin util.ProcessExecutionError | 16:37 |
magicalChicken | I'll figure out how to fix the formatting there for cloud-init, shouldn't be too difficult | 16:38 |
=== rtheis_ is now known as rtheis | ||
=== JasonF is now known as JayF | ||
harlowja | @smoser if u get a chance https://code.launchpad.net/~harlowja/cloud-init/+git/cloud-init/+merge/304001 | 20:11 |
harlowja | that i think should fix the 2.6 occurence that i think is busted | 20:11 |
harlowja | or at least one of them, lol | 20:11 |
harlowja | though i think there are more, lol | 20:13 |
harlowja | ie https://gist.github.com/harlowja/4ee5f7da70e422d918f3bfa8a5f2bf33 | 20:14 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!